Understanding the New AI Laws in California

How 19 new laws in California may impact AI for better or for worse.

San Francisco Bay Area, California (source)

“While well-intentioned, SB 1047 does not take into account whether an AI system is deployed in high-risk environments, involves critical decision-making or the use of sensitive data. Instead, the bill applies stringent standards to even the most basic functions — so long as a large system deploys it. I do not believe this is the best approach to protecting the public from real threats posed by the technology.”

- From Governor Newsom’s veto message of SB-1047

Introduction 

September was a productive month for the office of California's governor Gary Newsom which considered no less than 38 AI-related bills. Of these, 19 new bills with consequences for the development and deployment of AI were signed into law.  

To the celebration of AI enthusiasts and the sorrow of AI doomers, the sweeping SB 1047 was vetoed by Newsom’s office. Earlier in May, the bill had passed in California’s State Senate by 31 votes to 1. I was already preparing to write a deep dive on SB 1047 but now it doesn’t matter much. 

SB 1047 - which went under the full name “The Safe and Secure Innovation for Frontier Artificial Intelligence Models Act” – was intended to reduce the risks of “catastrophic harms” from AI and introduced numerous obligations to providers of the most powerful and expensive AI models. The bill also raised the blood pressure of many tech executives and venture capitalists in California and elsewhere due to its potentially damaging effect on innovation in AI. 

Now, instead of a broad legal framework that targets foundation model developers (foundation models are “models that are trained on broad data at scale and are adaptable to a wide range of downstream tasks”), California has opted for a subset of laws that each address specific applications of AI that are known to be harmful such as deepfakes in political disinformation campaigns, irresponsible use of generative AI in healthcare, "robocalls", and AI-generated nudes. 

In this post, we will run through the new laws which I have divided into four categories:

• Deepfakes and misinformation

• The rights of artists and creators

• The integrity of the human body

• Use of AI in the public sector

Although the bundle of new AI laws only directly impacts California, it will have far-reaching second-order effects and it will impact how AI models are used and developed in general, much like the AI Act in Europe.

Most of the world’s leading AI companies reside in California and with the state’s population of nearly 40 million people, it’s by no means an insignificant market. Additionally, the new rules in California may serve as an inspiration for other states in the US and the rest of the world. This is why I find the topic important to cover.